Trust in Numbers: The Pursuit of Objectivity in Science and Public Life

By Theodore M. Porter

A foundational work on historical and social studies of quantification

What accounts for the prestige of quantitative methods? The usual answer is that quantification is desirable in social investigation as a result of its successes in science. Trust in Numbers questions whether such success in the study of stars, molecules, or cells should be an attractive model for research on human societies, and examines why the natural sciences are highly quantitative in the first place. Theodore Porter argues that a better understanding of the attractions of quantification in business, government, and social research brings a fresh perspective to its role in psychology, physics, and medicine. Quantitative rigor is not inherent in science but arises from political and social pressures, and objectivity derives its impetus from cultural contexts. In a new preface, the author sheds light on the current infatuation with quantitative methods, particularly at the intersection of science and bureaucracy.

• Language: English
• Publisher: Princeton University Press
• Release date: Aug 18, 2020
• ISBN: 9780691210544

Book preview

TRUST IN NUMBERS

TRUST IN NUMBERS

THE PURSUIT OF OBJECTIVITY IN SCIENCE AND PUBLIC LIFE

Theodore M. Porter

New Edition

With a new preface by the author

PRINCETON UNIVERSITY PRESS PRINCETON, NEW JERSEY

Copyright © 1995 by Princeton University Press

New preface by the author copyright © 2020 by Princeton University Press

Published by Princeton University Press

41 William Street, Princeton, New Jersey 08540

In the United Kingdom: Princeton University Press

6 Oxford Street, Woodstock, Oxfordshire OX20 1TR

press.princeton.edu

Cover design by Karl Spurzem

All Rights Reserved

First cloth printing, 1995

First paperback printing, 1996

New paperback edition, 2020

New paperback ISBN 978-0-691-20841-1

Cloth ISBN 978-0-691-03776-9

Library of Congress Control Number: 2020936093

British Library Cataloging-in-Publication Data is available

This book has been composed in Electra

Printed in the United States of America

Contents

Preface to the New Edition

ALTHOUGH Trust in Numbers seems to work as a memorable title, it can mislead. The book is not about implicit trust, but reluctance and hesitation. Numbers that appear sufficiently routine may pass under the radar, but when conflicting interests are at stake, they are readily challenged. They often require standardization. This typically involves putting aside deep meanings and convictions in favor of compromise and convention. The title came to me in reaction to my editor’s suggestion of Truth in Numbers, which I rejected at once. Although an element of reverence for measurement and mathematics has long been with us, that is not what this book is about. It presumes no mystical disposition to put faith in figures, but features instead a push for impersonality of a sort that might even release us from the need for trust. The virtue of truth is less at issue here than the reliability that may come with constraint.

My title phrase, it turns out, has a history, and I was able to track it down for this edition with the aid of modern databases and search engines. Most prior uses of the phrase, it turns out, were disapproving. The word numbers usually referred not something abstract and mathematical, but to numbers of people. The earliest instance I have turned up, from a late-eighteenth-century epic poem, refers to the good fortune of the Athenians to face an enemy that cannot take advantage of superior numbers: What more / Can they solicit of propitious Heav’n, / Than such deluded enemies to face, / Who trust in numbers, yet provoke the fight / Where multitude is fruitless?¹ If this phrase was new, its disparagement of numbers was old, and its author, Richard Glover, may also have been thinking of King David’s forbidden and futile census. It was futile because the outcome of battle must be determined by God’s favor, not by a head count. A biblical guide by a learned New York rabbi summed up its lesson about 1890 in a chapter heading: Put not your trust in numbers. About the same time, a character in a novel, The Crusaders, about the first years of the women’s temperance movement, was made to declare: We must not trust in numbers, but trust in God. Finally, in the next century, guidance by numbers was becoming respectable in business management. Yet blind trust in numbers, even in the business writing, would be going too far. The preface to a 1913 textbook on business administration warns against succumbing to fads. Some business writers insist on large units, it explained, others on small ones. Some advocate extreme division, others extreme concentration of authority. We hear of trust in numbers and trust in leadership. Numbers might be useful, but there is no royal road.²

Since numbers are our topic, we may well ask: What do the numbers tell us about Trust in Numbers? A few clicks bring us to Google Scholar’s bar graph, which, I am pleased to report, shows a steady increase of annual citations, beginning with an easily surpassed zero in 1995, the year of publication. Well, to be strictly accurate, there are a few annual downturns, but none that can’t be dissolved into a moving average. The annual citation number first reached a hundred in 2001, then two hundred in 2008, and in 2017 surpassed five hundred, where it now lingers. The cumulative number passed six thousand in March 2020. But do numbers like these count for anything? Should they?

It is not only number-nihilists who have doubts. Official citation metrics typically ignore citations beginning just a few years after initial publication. Should we discount old citations? Yes, if you think that every new finding is quickly surpassed. You can’t expect to linger on the cutting edge. There are also, however, compelling pragmatic reasons for short-termism. All kinds of data-based decisions that have to be made in real time draw on these numbers. Maybe it isn’t an imperative of research science but rather a hustling, data-obsessed bureaucracy that calls the shots on numbers.

Indeed, scientists as well as humanists often complain that their work is judged superficially, and joke about meaningless manipulations of publishing habits that can enhance their prospects for academic advancement. The basic purpose of these metrics, after all, is to simplify administration and to facilitate the assessment of persons and programs by people who understand none of it. Nowadays, many decisions are made automatically by computers, which, though highly complex, comprehend nothing at all. Yet almost any standard, however problematic, tends to pick up credibility with use. This is based on the data, the managers explain, echoing all those scientists who use data-driven as the standard of legitimate knowledge. What they of course mean, but may have neglected to say, is that the data should be reliable and suited to the purpose. For decisions that matter, algorithmic reasoning may not rise to the desired standard of wisdom.

Serious scientists recognize the problems, of course, in practice if not always in theory. Still, it is to the credit of humanists that we are less inclined than many scientists to let ourselves be hoist by this petard of data-drivenness. Once you acknowledge that the significance of books and articles, including scientific ones, does not in general reduce to instrument readings or a regression coefficient with error bars, it may seem wise to dig into older, more complex texts for insights on pressing new problems as well as gnarly old ones. I do not claim that citation counts and the like must be pointless. I even admit to feeling some satisfaction that quite a lot of readers have found and continue to find in this book something that matters for their own research and writing. Even if I have some fun here with strained, naive, and disingenuous efforts to dissolve disagreement into calculation, Trust in Numbers is in no way a broadside against numerical reasoning. But those who undertake to toss qualitative reasoning out the door are likely soon to be found sneaking it back in through the window.

Although my book does not proceed chronologically, I have written as a historian. It takes up a variety of topics that are linked by the form of the argument, emphasizing practical and administrative uses of calculation over academic ones. Almost everything takes place between about 1830 and 1960, and I do without any all-encompassing historical ruptures between rival forms of objectivity. My more theoretical positions are worked out by way of exemplary episodes, a few of them quite detailed. They instantiate the conditions that have stimulated a push for routines of calculation. Such pressure has become more and more widespread since the 1930s, initially, I suggest, in the United States. Science of a more academic sort was also caught up in such pressures. This history has a direction that reflects a growing resistance to public decisions grounded on administrative rank or expert judgment. Objectivity came more and more to mean proceeding rigorously and impersonally from data to calculation. Yet each episode has also its own dynamic. When experts submit to unthinking routines, this generally is not because they have lost faith in their capacity to decide appropriately but because their capacity for disinterestedness has been called into question. At the same time, the story involves the production of new forms of expertise based on new tools of calculation. Over time, quantitative methods have become more and more the basis of expertise.

OBJECTIVITY QUESTIONS

THIS book was written, published, and reviewed during an era of science wars. I am thinking in particular of the book Higher Superstition, a vitriolic attack on some leading authors linked to social and cultural studies of science, and the Sokal hoax, a text composed to mimic such studies but festooned with physics howlers that the editorial collective of the journal Social Text could not detect.³ Sokal’s ruse was marginally clever and marginally immoral, and proved rather less than it claimed, while Higher Superstition exposed, as I thought, some questionable reasoning, yet with a blunderbuss rather than a rapier, and denounced without comprehending some of the best and most creative scholarship in our field. My arguments depend utterly on a growing recognition of the fundamental role of skill, community, and face to face contact in science, revealing, as Steven Shapin particularly emphasized, that science is by no means effortlessly cosmopolitan, but profoundly local. It can, in some important respects, be made universal, but this depends on a range of social as well as material technologies, and the work of science remains in a way artificial even when it has an excellent claim to be described as true.⁴

The objectivity of science has traditionally been discussed as the basis of its validity, indeed truth. By 1990, I was coming to think of it as a feat of standardization, that is, of making the local universal. Bureaus of standards have an important role in this, as does the manufacture of increasingly precise (hence uniform) instruments. In the book I speak of the universality of science as in important respects an artifact, adding that it is by no means arbitrary. My purpose was neither to celebrate nor to denounce claims of scientific objectivity, and the book was not taken as hostile to science. The lone exception is a review essay in Science by a historian of physics, focused gloomily on decaying mores of truth and trust. This defender of intellectual morals, however, gave little evidence of having read beyond the preface, apart from citing page 197, which proved on inspection to be not from the eighth chapter of Trust in Numbers, but the first page of my paper in a collection, Rethinking Objectivity, also included in the review. The evidence of a few pages had apparently sufficed to finger me as one of the dreaded postmodernists! Readers who read can reasonably judge for themselves.⁵

Objectivity is more often preached than analyzed, and few discussions of it were relevant to my purposes. The exceptions to this rule were not notably postmodern. In the course of my work, I came to understand that my graduate adviser’s notorious book from 1960, The Edge of Objectivity, by then very much out of favor among historians of science, was anything but conventional in its treatment of that concept. Charles Gillispie condemned as futile every attempt to extract moral meaning from knowledge of nature. The engagingly idiosyncratic lives of his scientific protagonists served as counterpoint to the tragic aspect of the story, for this edge, as he termed it, was neither neutral nor benign, but a cruel, unsparing knife edge. It had advanced mercilessly from mechanics to chemistry to biology, cutting away every hope that natural knowledge might promote moral meaning. Romanticized nature, which he described as tumescent, even sexual, might be inspiring as poetry, but it was, he declared, the wrong way for science.⁶ Every step forward for science entailed a sacrifice of human meaning.

Gillispie dealt harshly with moralists and romantics, from Diderot and Goethe to Ernst Haeckel and Arthur Koestler, all of whom desired to make nature a basis of morality and to align it with our deepest longings. Proper science, however, was abstract and intellectual, and the most technically capable scientists gave little thought to doing good. Even the crafts, he supposed, were far removed from abstract science. Opposed to cool science, the antiscientific yearning to inhabit or to merge with nature became dangerous in times of revolution. He meant in particular the 1790s, a focus of his scholarship, and the 1960s, through which he lived, discontentedly. In the wake of the latter experience, he worried that its spirit could tarnish the field of scholarship that he had helped to create. Yet he seems to have identified in some way with what he condemned, notably in the Werther-like episode he loved to recount when, caught up in currents of emotion as he looked down at the Charles River, he flung his slide rule into the waters to seal his decision for Harvard and a career in history over chemistry at MIT. Chemistry, he said, had been his duty, especially during the war. History was his passion.

Historians of science, many of whom abandoned the prospect of careers in science despite being quite good at it, are perhaps especially conscious of the sacrifice that science may entail. The focus and self-restraint of scientific objectivity, which can be severe, is one dimension of this sacrifice. Thomas Kuhn, another of my graduate teachers, argued at length in The Structure of Scientific Revolutions that it is defined and limited by the unsparing routines of normal science. For him, however, a life of mopping up was intolerable. He wanted a role in something great, a revolution, even at the cost being for a time without a paradigm. Eventually he found his revolution, not in physics, where he was trained, but in philosophical history.

THE PROBABILISTIC REVOLUTION

IN 1982, my research on the history of statistical sciences intersected with this theme of scientific revolutions. Kuhn’s brilliant book, The Structure of Scientific Revolutions, had given inspiration to Lorenz Krüge, the philosophical organizer of a research project at the Center for Interdisciplinary Research of the University of Bielefeld in Germany. It was to be called The Probabilistic Revolution, 1800–1930. The choice of probability as theme of the project owed much to Ian Hacking, who was then a Stanford professor, known particularly as a superbly original philosopher of probability. The big idea, as it took shape, was that probability theory had provided the basis for a wide-ranging, long-term revolution across many fields of knowledge, including philosophy and the social sciences. Such a revolution would be much wider in scope than those theorized by Kuhn. The Bielefeld project worked out very well, even if the probabilistic revolution itself proved an elusive quarry. Krüger’s vision pointed to a shift in worldviews from determinism to indeterminism, exemplified by quantum physics, and the resulting two-volume collection largely preserved this emphasis.

Six of the project participants, including me, subsequently took on the task of writing a more synthetic account, which appeared not long afterward as The Empire of Chance (1989). The interests of this subgroup focused more on statistical reasoning and the expanding role of numbers. Our topic also expanded to what we referred to in the subtitle as science and everyday life. In our discussion of potential unifying themes, we debated for a time the idea that the exclusion of human judgment in favor of rules might be a basic feature of science. Although most of us were not convinced, this push for rules did appear integral to statistical inference and calculation.⁷ Was scientific objectivity about constraints of this kind? The possibility seemed to open up a fresh domain of topics and questions. Our book in the end was mainly focused on themes growing out of the prior work of the six contributors but went beyond them in trying to make sense of the role of probability as a basis of expertise. We took up a mélange of topics: baseball, law, clinical trials, mental testing, insurance, judgment under uncertainty, and extrasensory perception. What really opened this topic up for me was my realization that the preoccupation with statistical objectivity in these heterogeneous domains could in no way be reduced to science envy. The extraordinary expansion of inferential statistics had not originated in physical science but rather in psychology, medicine, economics, ecology, eugenics, agriculture, and all kinds of practical and professional issues where public choices had to be made and defended.

Already in my first book, The Rise of Statistical Thinking, I was absorbed by the role of social and administrative numbers in the shaping statistical reasoning. We might speak more broadly of bureaucratic knowledge. In those days, historians of science hardly ever sank to this level of mundaneness. I was becoming convinced that the key sites of quantification were not at all limited to academic science and that the expanding role of quantification in social fields was in many respects continuous with governmental and commercial numbers.

This work was allied to larger trends. Inspired by emerging topics in research on technology, historians of science in those years were rediscovering how much scientific work took place away from the supposed ivory tower of academe. This point applied still more compellingly to the social sciences, which had little idea of scholarly detachment before the 1890s. Long before quantification achieved the status of a universal scientific ideal, it already was widely dispersed in practical life. Merchants, bankers, farmers, tax officials, and militaries, to give just a few examples, could scarcely function without counting and measuring. They also took a keen interest in standardizing materials and measures. Few things were more irritating to peasant farmers than the lack of fixity of the bushel measure, unless it was the rules that governed the heaping and compacting of grain. These controversies, and the political ideals associated with them, had a role in launching the metric system in the first years of the French Revolution. In this matter, physicists responded to peasant politics, though not in a way they found helpful.⁸

Scholarly sources and interpretations on the many forms of practical quantification were ubiquitous, once I knew how to look: histories of business information and agricultural measurement, the sociology and history of accounting, history of economics and econometrics, legal studies of regulation, political histories of expertise, historical and social studies of census practices, studies of the standardization of units of measurement, life insurance, the planning of public works, pharmaceutical trials, and much else. The topics we addressed in The Empire of Chance were mere selected tips in fleets of icebergs. The history of technology and the emerging field of science and technology studies offered better models and a more active engagement than mainstream history of science. And for my topic, English-language scholarship was far from dominant. Historical and sociological studies of statistics were especially lively in France, where they were actively pursued within the French national statistics office, INSEE, and state engineering bodies as well as academic centers like the graduate school of social sciences, EHESS. These were, however, highly diverse traditions, and my colleagues there appeared a bit nonplussed when I jetted in from California and (in accented French) spoke as if their diverse researches were all part of a single package.

Indeed, there were important contrasts among these heterogeneous scholars I was reading. The elements they shared were not confined to a single country or language. What unified them was a sense of numbers, statistics, and calculation as instruments of objectivity. Numbers could be deployed, for example, as Alain Desrosières put it, to make things which hold together. Even when the results of these calculations remain vulnerable to challenge, if legal authorities are yet willing to grant that the rules have been properly followed, that may be objectivity enough. And when they fail, the usual remedy is not to disassemble this machinery of standardization but to reconcile contradictions by extending it. What appears most dull may conceal wonderful secrets. When everything appears routine and boring, success is assuredly within reach.⁹

NUMBERS PAST AND PRESENT

BACK when this book first came out, I wondered if the need for formulaic assurance of objectivity might have peaked. Behind this idea lay a hypothesis: that as the sciences most dependent on statistical tests gained more confidence in their own disciplinary practices, the pressure to conform to rules of calculation could be relaxed. This anticipation has proved to be quite wrong. The collapse of my prophecy, however, has been good news for my book, as its citation trajectory demonstrates. Statistics, as a discipline, is currently expanding into data science, while data, algorithms, and indicators are by now making a good show of ruling the world. Quantitative dreams of effortless regulation and disarmed mistrust still proliferate at the intersection of academic and bureaucratic life. Statistical routines have been put to work in therapeutic medicine, studies of classroom effectiveness, policing, development economics, and ten thousand rankings and metrics. All have been seriously criticized, yet they continue to be marketed, indeed ubiquitously, as the holy of holies. Is expertise giving way to algorithms?

Two of the longest chapters in Trust in Numbers, and the ones that seem to have made the greatest impression, concern a contrast between nineteenth-century French engineers and twentieth-century American ones on the evaluation of construction projects. Both came to involve quite a lot of economic expertise as a basis for these political choices, and both tried to make the politics invisible. In the end, the French were better able to defend their expertise as a basis for correct decisions, while the less unified American engineers repeatedly lost control of the politics and were more or less compelled to proceed as if their choices were made strictly by the numbers. But that meant they could be held to account for inconsistencies in their quantitative practices, and in this way the pretense of subordinating judgment to objective cost-benefit analysis began to be made real. Before long, reliance on numerical data made its appearance in social science as the very model of rational decision making.

In academic life, decision by calculation usually comes down to a statistical test of significance and a p-value, giving the probability that an improved harvest in fertilized plots or better health outcomes for patients given an experimental drug could appear simply by chance. These tests are now much criticized for focusing on the wrong question or leaving too much room for abuses. In this context, citation counts have also generated controversy. There are many good reasons not to apply them automatically as the basis for rewards and demerits, but such restraint becomes more and more difficult as they congeal into systems. They extend now over multiple levels of routinized measurement. Measures of research merit, for example, begin simply enough with publication lists. Why not count them? But as a result, the least publishable unit acquires an irresistible appeal for authors of scientific papers. Where you once had just one publication, now you have several. It now becomes more urgent to add a measure of research quality. Citations, for example. In pre-electronic days, these were compiled into bound volumes and printed every few months. Electronic publishing simplified such tallies enormously. Indeed, it became possible to assign scores to journals as well as to scientists and their papers. Meanwhile, and not wholly by coincidence, academic publishers entered on a path of extreme profiteering. Libraries now rely on impact factors to decide which journals to add and which to cancel. With impact factors, also, a new paper can be scored immediately, based on an average for the journal, rather than having to wait until citations came in. Promotions and salaries already depended on numbers like these as well as on revenue from research grants (which in turn entail the expectation of publication in refereed, high-impact journals). By now, the stakes are high for everyone involved, and everyone knows something about how the system can be gamed. All the while, the science of bibliometrics is groping for better ways to reduce two numbers, publications and citations, to just one combined number such as the famous h-index. For whom is two too many?! All these numbers provide a basis for ranking academic departments, schools, and universities. Philanthropists, who are not always charitable, rely on such numbers as they consider the prospective recipients of their largesse. The train of quantitative evaluation has definitely left the station, and it is not easy to go back. Of course, it is not just universities. A similar logic applies to hospitals, prisons, and much more. Trust in numbers, in this context, is trust in a very specific sense, implying often a radical distrust of people and institutions.¹⁰

Some of the liveliest and most consequential debates about metrics are now focused on schools and teachers, whose work can be conveniently but imperfectly measured using standardized tests. The comparisons are often criticized as unfair, since it is so hard to correct for population-level differences such as family backgrounds. Also, standard tests make little sense if schools pursue subtly or fundamentally different educational goals, such as vocational versus academic. In the United States, schools have been shuttered on the basis of low test scores. Internationally, the Program for International Student Assessment, or PISA, defines a basis for comparison at the level of the country, despite multiple dimensions of radical heterogeneity. These measures have provoked extensive discussion and debate about their validity as well as their curricular implications.

The analysis or manipulation of numbers involves a special expertise, one that can undermine or override alternative forms of knowledge. The world is made thinner when subtle reasoning is devalued for the sake of administrative convenience. Financial and information technologies flourish in this kind of world, since the ambiguities provide opportunity for manipulation. Yet there are critics, too, many with the sagacity to appreciate the stakes of statistics. Trust in Numbers has endured and even, in its modest way, flourished in the context of debates like these, as a voice of measured skepticism. Its readers seem to perceive how history can serve as a basis for engaging with present concerns. Such issues matter especially now to professionals, who find their expertise giving way to standardized numbers that can foster undesirable incentives. I draw back here from simply blaming bureaucracy, which is rarely autonomous and which admits a variety of forms. It also can function as an important repository of knowledge and expertise. Bureaucracy, too, can be reshaped and flattened by distrustful trust in numbers. While it is not for experts and professionals to run the world, their knowledge should be heard and taken seriously on the matters they know best.¹¹

Notes

1 Richard Glover, The Athenaid, a Poem, 2 vols. (Dublin: Printed for Messrs. Byme, M’Kenzie and Moore, 1788), vol. 1, 85 (book 5, lines 258–262). The earliest appearance of the phrase trust in numbers given by HathiTrust is 1831, but it missed one essential letter from the first edition of Glover’s poem.

2 Maurice H. Harris, The People of the Book, 3 vols., 7th ed. (New York: Published by the author, 1897), vol. 2, 201–204; Emma R. Norton, The Crusaders: A Story of the Women’s Temperance Movement of 1873–74 (New York; Peabody, Macey & Co., 1892), 27; Edward W. Jones, Business Administration: The Scientific Principles of a New Profession (New York: Engineering Management Co., 1913), 1–2.

3 Paul Gross and Norman Levitt, Higher Superstition: The Academic Left and Its Quarrels with Science (Baltimore, Md.: Johns Hopkins University Press, 1994); Alan Sokal, Experiments with Cultural Studies, Lingua Franca, 4 (May 1996)

4 This line of argument is most memorably articulated in Steven Shapin and Simon Schaffer, Leviathan and the Air-Pump (Princeton, N.J.: Princeton University Press, 1985), 1.

5 Paul Forman, Book Reviews, Science, 269 (August 4, 1995), 706–710, on 709; Theodore M. Porter, Objectivity as Standardization: The Rhetoric of Impersonality in Measurement, Statistics, and Cost-Benefit Analysis, in Allan Megill, ed., Rethinking Objectivity (Durham, N.C.: Duke University Press, [1992] 1994), 197–237; Rob Hagendijk, An Agenda for STS: Porter on Trust and Quantification in Science, Politics, and Society, Social Studies of Science, 29 (1999), 629–637, is an excellent review of the reviews of Trust in Numbers.

6 Charles Coulston Gillispie, The Edge of Objectivity (Princeton, N.J.: Princeton University Press, 1980), 201.

7 Gerd Gigerenzer, Zeno Swijtinki, Theodore Porter, Lorraine Daston, John Beatty, and Lorenz Krüger, eds., The Empire of Chance: How Probability Changed Science and Everyday Life (Cambridge: Cambridge University Press, 1989). The original collection is Lorenz Krüger et al., eds., The Probabilistic Revolution, 2 vols. (Cambridge, Mass.: MIT Press, 1987).

8 On quantification and the social sciences, see my essays and others in Theodore M. Porter and Dorothy Ross, eds., The Cambridge History of Science, vol. 7: Modern Social Sciences (Cambridge: Cambridge University Press, 2001); Theodore M. Porter, The Social Sciences, in David Cahan, ed., From Natural Philosophy to the Sciences: Writing the History of Nineteenth-Century Science (Chicago: University of Chicago Press, 2003), 254–290; also Theodore M. Porter, The Rise of Statistical Thinking, 1820–1900 (Princeton, N.J.: Princeton University Press, 1986). On grain measuring practices, see Witold Kula, Measures and Men, trans. Richard Szreter (Princeton, N.J.: Princeton University Press, 1986); on origins of the metric system, Ken Alder, A Revolution to Measure: The Political Economy of the Metric System in France, in M. Norton Wise, ed., The Values of Precision (Princeton, NJ: Princeton University Press, 1995), 39-71).

9 Alain Desrosières, How to Make Things Which Hold Together: Social Science, Statistics, and the State, in P. Wagner, B. Wittrock, and R. Whitley, eds., Discourses on Society, Sociology of the Sciences Yearbook, vol. 15 (1990), 195–218.

10 On rankings, see especially Wendy Espeland and Michael Sauder, Engines of Anxiety: Academic Rankings, Reputation, and Accountability (New York: Russell Sage, 2016).

11 Theodore M. Porter, Thin Description: Surface and Depth in Science and Science Studies, Osiris, 27 (2012), 209–226; Richard Rottenburg et al., eds., The World of Indicators: The Making of Governmental Knowledge through Quantification (Cambridge: Cambridge University Press, 2015).

Preface

SCIENCE is commonly regarded these days with a mixture of admiration and fear. Until very recently, though, English-language historians of science were more likely to resent its pretensions than to fear its power. Here resentment grew out of reverence. Karl Popper and Alexandre Koyré, who gave form to brilliant traditions in the philosophy and history of science beginning especially in the 1950s, agreed that science was about ideas and theories. Koyré gave priority to thought experiments over the work of hands and instruments, and wondered, famously, if Galileo had ever performed any experiments at all. Popper allowed that experimentation could falsify theories, but held that the real work was done when the theory was adequately articulated. Experimenters had no more than to carry out what the theory dictated. Both praised science as a model of intellectual and philosophical achievement. Neither provided any reason for thinking that science could have much to do with technology. Still less could the hierarchical imagination of the historian or philosopher of science conceive that social science was authentically powerful.

This problem of the relations of science to technology inspired nothing like the heated (and, it now seems, empty and incoherent) controversy over the relative merits of externalist and internalist explanations of scientific change. Rather than arguing, much of the profession took for granted that science had the loosest connections with the practical world of engineering, production, and administration. In retrospect, I can see that my graduate training provided ample opportunity to form a more judicious view. My teachers learned earlier than I did to appreciate the limitations of seeing the scientific enterprise mainly as a pursuit of theory. Still, I think I was not unusual among historians of science of my generation in thinking that the widespread linking of science and technology or of science and administrative expertise involved something fundamentally spurious, that these supposed connections brought undeserved credit to each enterprise by making science seem more practical and its applications more intellectual than either really is.

A critique of this nature underlay my original formulation of this project. I planned to examine the history of neoclassical economics, the most mathematical of social science disciplines—indeed, possibly the most mathematical of all disciplines. Economics values most highly this supremely abstract mathematics, yet somehow economists sustain the image of a discipline capable of telling businesses and governments how to manage their affairs more effectively. I expected to show through an analysis of the relations of economics to policy that academic economics was a kind of sport, empty of implications for economic practice.

That is not the book I have written. It didn’t take long to realize that neoclassical economics has had many critics who were better informed than I was likely to become. I found also that the economics discipline involves a greater variety of tools, aims, and practices than I had appreciated, and while I still think there is need for a more profound consideration of the relations between economic mathematics and the practices that support forecasting and policy advice, I am not the one to undertake it. In any case, my earlier suspicion that mathematics and policy were almost independent worked badly as a way of formulating a historical project. Its validity was even more damaging than its shortcomings. If, indeed, neoclassical mathematics is irrelevant to the economic world, my history of the relations between economics and policy would turn into the history of nothing at all.

So I have taken here a different tack. The interpenetration of science and technology, I now concede, is unmistakable, especially in the current century. That of social knowledge and social policy is only slightly less so. How are we to account for the prestige and power of quantitative methods in the modern world? The usual answer, given by apologists and critics alike, is that quantification became a desideratum of social and economic investigation as a result of its successes in the study of nature. I am not content with this answer. It is not quite empty, but it begs some crucial questions. Why should the kind of success achieved in the study of stars, molecules, or cells have come to seem an attractive model for research on human societies? And, indeed, how should we understand the near ubiquity of quantification in the sciences of nature? I intend this book to display the advantages of pointing the arrow of explanation in the opposite direction. When we begin to comprehend the overwhelming appeal of quantification in business, government, and social research, we will also have learned something new about its role in physical chemistry and ecology.

My approach here is to regard numbers, graphs, and formulas first of all as strategies of communication. They are intimately bound up with forms of community, and hence also with the social identity of the researchers. To argue this way does not imply that they have no validity in relation to the objects they describe, or that science could do just as well without them. The first assertion is plainly wrong, while the latter is absurd or meaningless. Yet only a very small proportion of the numbers and quantitative expressions loose in the world today make any pretense of embodying laws of nature, or even of providing complete and accurate descriptions of the external world. They are printed to convey results in a familiar, standardized form, or to explain how a piece of work was done in a way that can be understood far away. They conveniently summarize a multitude of complex events and transactions. Vernacular languages are also available for communication. What is special about the language of quantity?

My summary answer to this crucial question is that quantification is a technology of distance. The language of mathematics is highly structured and rule-bound. It exacts a severe discipline from its users, a discipline that is very nearly uniform over most of the globe. That discipline did not come automatically, and to some degree it is the aspiration to a severe discipline, especially in education, that has given shape to modern mathematics.¹ Also, the rigor and uniformity of quantitative technique often nearly disappear in relatively private or informal settings. In public and scientific uses, though, mathematics (even more, perhaps, than law) has long been almost synonymous with rigor and universality. Since the rules for collecting and manipulating numbers are widely shared, they can easily be transported across oceans and continents and used to coordinate activities or settle disputes. Perhaps most crucially, reliance on numbers and quantitative manipulation minimizes the need for intimate knowledge and personal trust. Quantification is well suited for communication that goes beyond the boundaries of locality and community. A highly disciplined discourse helps to produce knowledge independent of the particular people who make it.

This last phrase points to my working definition of objectivity. It is, from the philosophical standpoint, a weak definition. It implies nothing about truth to nature. It has more to do with the exclusion of judgment, the struggle against subjectivity. This impersonality has long been taken to be one of the hallmarks of science. My work broadly supports that identification and tends to the view that this, more than anything else, accounts for the authority of scientific pronouncements in contemporary political life. Once again, though, I am reluctant to make science the unmoved mover in this drive for objectivity. In science, as in political and administrative affairs, objectivity names a set of strategies for dealing with distance and distrust. If the laboratory, like the old-regime village, is the site of personal knowledge, the discipline, like the centralized state, depends on a more public form of knowing and communicating. Quantification is preeminent among the means by which science has been constructed as a global network rather than merely a collection of local research communities.

Some of the best and most fashionable recent work in science studies has aimed to understand science as a thoroughly local phenomenon. The genre of microhistory, which has enjoyed brilliant success in cultural history, has become influential also in the history of science. I have learned a great deal from this work, and I hope I have adequately appreciated its virtues. It provides a superb point of departure for studies of science, precisely because it renders the universality of scientific knowledge problematical. But it does not simply negate it. Science has, after all, been remarkably successful at pressing universal claims and gaining international acceptance. Explaining this achievement, and unpacking its implications, ought to be central problems of the history of science. The account I give here is mainly cultural and, broadly, political. I suggest that the problems of organization and communication faced by science are analogous to those of the modern political order. This is not meant to imply that science is not constrained in important ways by the properties of natural objects, nor even that the forms of language and practice I discuss are independent of those properties. I do not claim that quantification is nothing but a political solution to a political problem. But that is surely one of the things that it is, and our understanding of it is poor indeed if we do not relate it to the forms of community in which it flourishes.

The argument, as I have presented it so far, is as much sociological or even philosophical as historical. Since I am unlicensed in both the former domains, I tremble at the thought of writing a book that is not securely historical. The flow of topics and arguments in the book, however, is hard to reconcile with narrative or analytical history. Indeed, the book does not conform well to any established genre of scholarly writing. But there is, I like to think, some method to this madness. I should perhaps explain at the outset the pressures and strategies that have given shape to this study.

I began, as I have already explained, with the intention of studying the modern history of social quantification in relation to academic disciplines. Soon I found myself paying more attention